Re: Future XSLT expansion.

[Paul Tchistopolskii <paul@xxxxxxx>]

> Please stop the pointless comparisons to perl, and stop quoting each
> message in full.
 
> You said
> > Sounds amazing.
> 
> followed by yet another diatribe against the document function.
 
That is not true, what you are writing here. If quoting my letter you'l 
see that there was no diatribe against the 'document()' function, 
but there was the translation  of some 'advice' , given by XML 
specialist. 

Unfortunately, because in your last letters you ended saying that:
"you anyway have that node-set typecast"  :

----- Original Message ----- 
From: David Carlisle <davidc@xxxxxxxxx>
To: <xsl-list@xxxxxxxxxxxxxxxx>
Sent: Sunday, March 19, 2000 1:41 PM
Subject: Re: xt:node-set usage (was Re: Future XSLT expansion.)


> 
> > Because of this I can only use core XSLT functions that are required
> > to be in all XSLT implementations.
> 
> although in this case xt saxon xalan msxml all offer this functionality
> (MS not quite legally) and it is possible within a conforming stylesheet
> to ask if the function is available. So it is probably hard to find an
> XSL system where you can not coerce a result tree to a node set, and 
> you can still do it while writing a valid xslt stylesheet.
> 
> David
>  

That means we have no reasonable and honest discussion here, 
but we have something else. 

I'm not interested in spending my life explaning differences between : 
"you could use xt:node-set" and "node-set typecast is in the core", 
because I don't think this obvious difference should be explained 
even to student.
 
> > Because there is no typecast in the core, but because there is 
> > one special ( we know the secret ) hack which has a built-in 
> > typecast ( it is document() hack ) - if you want some variable to be 
> > converted to node-set - just place that variable into separate file.
> 
> I can't imagine what features the above is supposed to be describing.
> whether the relative url in document points to the same or different
> file has no effect on the result.
> 
> If you object to document() you should presumably object to the fact
> that the top level input file is also provided to the stylesheet as a
> node set? The semantics appear to be identical.

The answer to this 'argument' is already in the thread.
See below.

Sorry for quoting the entire letter. The letter below also shows why 
MS did better job here than all the 'conformant' engines.

MS allows natural on-the-fly  node-set typecast instead of  'document'
hack.

Those who are bashing MS, should start thinking about the real level of 
engeneering expertize in MS and in some *L-list. 

I'm not more participating in this thread, because I think everything 
is clear. 

I not only explained the problem in XSLT core , but I also wrote a set of 
proposals how to fix the problem in XSLT core, making XSLT more 
powerful, scalable and reasonable than it is now.

The proposal was to bring the node-set typecast into the core
and than rethinking the semantics of all the functions returning 
node-sets and in particular document() function.

I think any 'question' is already answered by the thread. 

If anyone has any problem understanding what the heck this is 
all about - you are welcome to write me privately and I'l answer 
what letters to re-read.

If anyone has an impression that I don't understand some
place in the standard - I suggest to think twice and to re-read 
the entire thread. 

If after that operation, the 'contradiction'  will not disappear, 
I'l be glad to get a reasonable criticizm, because probably 
there *is* some reason in current XSLT self-restriction 
( that means MS realy has some problem ;-). 

 So far there was no reasons given except usual XML-ish:

" It is in W3C paper! It is the STANDARD, you, fool mortal ! ".

Rgds.Paul.

----- Original Message ----- 
From: Jon Smirl <jonsmirl@xxxxxxxxxxxx>
To: <xsl-list@xxxxxxxxxxxxxxxx>
Sent: Saturday, March 18, 2000 10:22 AM
Subject: Re: Future XSLT expansion.

> Am I experiencing a side effect of this discussion?
> 
> I would like to do this:
> 
> <xsl:variable name="foo">
>     <mytag>
>         data
>     </mytag>
> </xsl:variable>
> 
> <xsl:apply-templates select="$foo"/>
> 
> This gives me a "cannot convert to node-set error" with XT.
> It was my understanding that this was in XSLT to prevent
> feedback from the resultset back into the input of the pattern
> matches.
> 
> Instead I use:
> 
> <mydata>
>     <mytag>
>         data
>     </mytag>
> </mydata>
> 
> <xsl:apply-templates select="document('')/mydata"/>
> 
> The downside to this is it causes my source XSL stylesheet to
> be parsed twice since XML and XSLT have different white
> space rules.
> 
> BTW - the Release 3 of the Microsoft XML dll allows the
> first case so it is in disagreement with JClark's XT.
> 



 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list